Pay-to-Quit Incentives in Teams

2026-07-15

2027-07-15

The primary outcomes are: (1) voluntary turnover — whether a team member accepts the offer to leave the team (quit vs. stay) at each decision point, and the team-level turnover rate; and (2) team performance — team output per production round, measured as the sum of members' scores (low task interdependence) or the highest single member's score (high task interdependence).

Secondary outcomes include: individual performance (number of items correctly decoded per round); helping behavior (the number and proportion of a member's correct decodes contributed to teammates' scores rather than their own); and self-reported team attachment, coordination, and cohesion measured in a post-experiment questionnaire.

Participants recruited through Prolific complete a real-effort decoding task in three-person teams over a series of timed production rounds, earning experimental currency that is later converted to U.S. dollars. During the study, team members may have the opportunity to voluntarily leave their team. When a member leaves, a participant drawn from a waiting pool joins so that team size stays constant. The study examines how features of the team's compensation and work environment relate to members' decisions to stay or leave, individual and team performance, and cooperation among teammates.

Not available

Randomization is performed by computer using the experimental software (oTree). Scheduled Prolific sessions are assigned to experimental conditions, and within each session the software groups participants into three-person teams.

Experimental session (assigned to one of the four conditions). Participants are randomly grouped into three-person teams nested within sessions. The team is the unit within which pay and performance outcomes are jointly determined.

Yes

120 teams

360 main team participants (120 teams × 3 members). An additional 300 (approx.) replacement-pool participants may join teams over the course of the study, for approximately 660 participants total.

30 teams per condition

All power analyses use G*Power (α = .05, power = 0.80). The design is a 2×2 between-participants experiment with 30 teams per cell (120 teams; 360 main team participants). Team performance: For the PTQ × Task Interdependence interaction on team output, the minimum detectable effect is Cohen's f = 0.26 (a medium effect). Team output is measured in decodes. Using the team-performance data in Shi, Tafkov, and Zhou (2025) to estimate a within-cell standard deviation of 75 decodes, this corresponds to a minimum detectable difference-in-differences of approximately 78 decodes (Cohen's f [0.26] × 4 groups × SD [75]). Relative to the mean team output across the six cells of Shi et al.'s Table 1 (555.4 decodes), this represents approximately 14% of baseline team output. Turnover: Turnover is a binary, participant-level outcome (quit vs. stay). For the effect of PTQ on the quit rate (present vs. absent), comparing two groups of 180 participants (360 main participants total, collapsing across task interdependence), the minimum detectable effect is Cohen's h ≈ 0.21. Anchoring plausible quit rates to a related PTQ experiment, in which the observed switch rate was 22.0% in the period the incentive was introduced and 61.0% in the following period, this corresponds to a detectable difference of roughly 9 percentage points at a ~22% baseline (≈22% vs. ≈31%) and roughly 10 percentage points at a ~61% baseline. For the PTQ × Task Interdependence interaction on the quit rate, the minimum detectable effect is Cohen's f ≈ 0.15. Expressed as the difference-in-differences in quit rates, this corresponds to approximately 24 percentage points at a ~22% baseline quit rate and approximately 29 percentage points at a ~61% baseline. Notes. The turnover analyses treat the binary quit outcome under a linear-probability/ANOVA approximation; because a proportion's variance depends on its level, the percentage-point translations are anchored to representative baseline quit rates from a related PTQ study. None of the figures above yet adjust for clustering of quit decisions within teams. Because turnover decisions are positively correlated within a team, the design effect will exceed one, and the cluster-adjusted MDEs will be larger.